TScale – Distributed training on consumer GPUs

I&#x27;m trying to run this but fo.cpp doesn&#x27;t exist in the repository. I made an issue see <a href="https:&#x2F;&#x2F;github.com&#x2F;Foreseerr&#x2F;TScale&#x2F;issues&#x2F;1">https:&#x2F;&#x2F;github.com&#x2F;Foreseerr&#x2F;TScale&#x2F;issues&#x2F;1</a>

What is this 1T index technique they seem so hyped about?

Can the inference piece be partitioned over multiple hosts?<p>Edit: algorithmed or partitioned in a way that overcomes the network bottleneck

tbh i never get why people keep reinventing config parsers, but i guess old habits die slow

Interesting that you put code in code folder, not src.

Even with consumer GPUs, the AI stack is completely dependent on ASML, isn&#x27;t it?<p>Thought experiment: What would happen if the Dutch government decided that AI is bad for mankind and shuts down ASML? Would the world be stuck in terms of AI? For how long?